Oh Those Road Woes
11.01.04
The justification for adding a road-win bonus to the RPI is based on the premise that the home team wins 70% of college basketball games. This has been cited in nearly every article on the new RPI, and usually as a direct quote from basketball committee chairman Bob Bowlsby. This is a classic case of using stats in a misleading way. While that number is technically accurate, it is inaccurate to assume that all of those wins are due to a home-court advantage. For instance, in a large majority of non-conference games the home team also happens to be the better team, and in many cases the much better team. Did Illinois beat Maryland Eastern Shore because the game was played in Champaign?
One way to eliminate this effect is to limit the scope to conference games. This way ensures that the better team is equally at home and on the road in the sample. For all 2500+ conference games in the 2004-05 season, the home team won 62% of its games. But more to the point, let's find out how hard it is for teams being considered for an at-large bid.
Taking teams ranked 34 through 54 in last year's basic RPI, their difference in home and road winning percentage is striking. I chose this range because this includes teams generally considered on the bubble. It turned out in 2004, that 8 teams in this range got an at-large bid, 8 were left out, and the remaining 5 received automatic bids.
Overall these teams combined for a 89-84 (.514) record in conference road games. At home they were 135-38 (.780). It's interesting what happens when the at-large teams are separated from the NIT teams. The 8 at-larges went 39-27 (.591) on the road while the 8 NIT teams went 23-42(.354). Their respective home records were almost identical (51-15 vs. 47-18). Road play is clearly what separates the men from the boys. For these teams, it is not so hard to win on the road.
Of those 16 teams looking for an at-large bid in the slot I looked at, only two of the eight selected (Arizona and South Carolina) did not have a winning record in conference road games. Of the eight that were not picked, only Utah State had a winning road record in conference. Utah State was regarded as the last team left out of the field.
So it seems like a good idea to include site data in the RPI, except when you consider that the committee did a great job on its own of weeding out the folks that succeeded away from home. As I said last December:
The team that does the best job, year-in year-out, of abusing the home court advantage is Hawaii. It’s really no fault of theirs because they never have to schedule non-conference road games and they usually get some decent competition to come to the islands. This year they will have to head to the mainland to play in the Bracket Buster in February. But for the other 325 teams, the distribution of game locations doesn’t make enough difference to justify making a simple formula more complicated.
These posts were motivated by a discourse between Joe Lunardi and Jeff Sagarin last year in which Sagarin attempts to promote his own super-secret formula to replace the RPI. But college basketball does not need a super-secret formula - it hasn’t worked out real well for college football.
And considering its simplicity, the RPI is pretty good at what it's supposed to do. Winning games is good, playing a tough schedule is good. But more fundamentally, just being a good team is good. There’s really no easy way to cheat the system. And even if you could, there is a bunch of humans waiting at the end of the season to try to smooth out the problems.
Now that we know (sort-of) what the NCAA will do to adjust the RPI, just how much of a difference will it make? Was Notre Dame coach Mike Brey was right when he said,
"If we had that formula last year, we're a seven seed." [South Bend Tribune, 10/23/04]
Next Post: Last season's adjusted RPI.
It Doesn’t Matter
10.29.04
The announcement that the NCAA will track a new "adjusted RPI" has generated a little buzz recently. Seth Davis came out of summer hibernation to write about it, Andy Katz followed, Joe Lunardi has piled on some thoughts of his own, and now local beat writers are in the mix.
Here's the truth: this alternate formula that is designed to reward road play will have almost no impact on the selection process, and it will have absolutely no impact on scheduling. I'll post a few things over the next few days to prove this. But first I have to establish some basic points so we're all on the same page.
Today's issue is from the Davis column:
However, that ranking will be evaluated separately from the standard RPI, which is based 75 percent on who a team plays, and 0 percent on where.
While the RPI formula literally factors strength of schedule (SOS) as 75% of the final RPI value, this is not how it works in practice. I went over this in last year's defending the RPI series. You should read it if you get a chance because the content still applies. I'll repeat some points from it over the next few days because they will address a number of items related to an "adjusted RPI."
Since SOS values have less variance (they tend to hover closer to the .500 mark) than winning percentage, a team's movement in the RPI is slightly more dependent on winning percentage than SOS. To use numbers from 2004, the range of winning percentage last year for all 326 teams was 0.9310, and the range of SOS was 0.2417. If you multiply the 75% by .2417, the result is .1812. This is the maximum possible difference that can be achieved in the SOS part of the formula. For winning percentage, the difference is 25% times .9310 with the result being .2328. So in practice, winning percentage has more influence on RPI than schedule strength.
This is demonstrated by the fact that of the 23 teams that finished with 6 or fewer losses last season, only Mississippi Valley State failed to crack the top 100 of the RPI. This was due to the Delta Devils having a schedule strength ranked 315 out of 326 - about as low as possibly could be attained. Of the 22 teams that made the top 100, seven had an SOS in the bottom half of college basketball. While a great RPI needs a good record and a good schedule, a good RPI can be achieved simply by winning.
Next post: Is it really twice as hard to win a road game than a home game?
Not Fast Enough
05.10.04
Perhaps the most talked about aspect of the 2004 National Chamionship game was Georgia Tech's insistence on forcing a fast pace. It was widely regarded as a mistake for the Jackets to avoid creating a halfcourt game. I'm not exactly sure why, because the Jackets could match the Huskies in both athleticism and depth. But did the Jackets play too fast - and if so, was this a mistake?
Here are the pre-tourney Georgia Tech games that involved the fastest pace, based on calculated possessions, and how the Jackets fared:
@ North Carolina (91 possessions) L 103-88 vs. Connecticut @ MSG (84) W 77-61 vs. Duke (81) L 82-74 @ Cornell (81) W 90-69 vs. Maryland (80) W 81-71 vs. UNC (77) W 88-77
Now the slowest paced games...
vs. Duke @ Greensboro (64) L 85-71 @ Clemson (66) W 79-60 vs. NC St. (67) L 79-69 @ NC St. (67) L 76-72 @ Georgia (67) L 83-80 @ Wake Forest (68) W 73-66
It appears that the Jackets were more comfortable with the faster pace, even against elite competition. Their first meeting against UConn was the 2nd fastest paced game they played. Against Duke, their worst effort came in the slowest paced game of the 3 meetings, in the ACC tournament. Their win at Durham was the 7th fastest game they played.
Now let's check out how UConn fared in similar games.
The slowest...
@ Syracuse (60) L 67-56 @ Villanova (60) W 75-74 vs. Pittsburgh (61) W 61-58 vs. Yale (62) W 70-60 vs. West Virginia (63) W 88-58 vs. Notre Dame (63) W 61-50
The fastest...
vs. Iona (89) W 104-54 vs. Sacred Heart (85) W 111-64 vs. Ball State (84) W 101-62 vs. Georgia Tech @ MSG (84) L 77-61 @ North Carolina (83) L 86-83 @ Rice (82) W 92-83 vs. Oklahoma (78) W 86-59 vs. Nevada (77) W 93-79
Certainly UConn was uncomfortable with a deliberate pace. The Yale game raised red flags with all sorts of observers that the Huskies were undeserving of their preseason #1 ranking. The late-season loss to Syracuse and the close call against 'Nova were surprising. Yet the Huskies also were human at the frenetic pace, losing to both Georgia Tech and UNC, the 2 fastest games they played against decent competition.
All in all, I think the case that Georgia Tech should have been more patient is a weak one. I doubt Paul Hewitt even considered doing anything other that letting his team play its usual game. They proved they could beat UConn with their running style earlier in the season, and the trio of games against Duke demonstrated that a faster pace gave them a better chance of success. There was the sweep by NC State and the shocking loss against Georgia to demonstrate that a slower pace could lead to disaster.
The game turned into a disaster anyway from the Tech perspective. UConn was a far superior team that would have beaten Tech regardless of what style the Jackets wanted to play.
For the record, the championship game had 76 possessions, 8 fewer than the regular season matchup. So maybe Georgia Tech was too patient.